• By Mohsin shakoor
• August 21, 2025August 21, 2025
• Education

Quick environment setup (choose one)

Option A – venv (works everywhere)

create & activate

python -m venv ai-basics

Windows:

ai-basics\Scripts\activate

macOS/Linux:

source ai-basics/bin/activate

upgrade pip

macOS Apple Silicon tip: if TensorFlow fails, try pip install tensorflow-macos and (optional GPU) pip install tensorflow-metal.

3) TensorFlow: tiny “learn y = 2x – 1” demo

What you’ll learn

• Build a minimal Keras model
• Train with gradient descent
• Make a prediction

print(“CUDA available?”, torch.cuda.is_available())

5) scikit-learn: quick classification (Iris + Logistic Regression)

What you’ll learn

• Train/test split
• Fit a model
• Evaluate accuracy + simple report

6) When to use which?

• TensorFlow/Keras: production pipelines, mobile (TF Lite), easy high-level APIs.
• PyTorch: research feel, very pythonic, flexible training loops, huge community.
• scikit-learn: classic ML (tabular data), fast baselines, preprocessing, metrics.

7) Common pitfalls & quick fixes

• Install errors → upgrade pip: python -m pip install --upgrade pip
• PyTorch GPU mismatch → use CPU wheels (command above) first; add CUDA later.
• TensorFlow on Apple Silicon → tensorflow-macos (+ tensorflow-metal for acceleration).
• Notebook kernel not seeing packages → install packages inside the same environment your Jupyter kernel uses (or pip install ipykernel; python -m ipykernel install --user --name ai-basics).

• Share
• Share
• Pin